[mirror_ubuntu-zesty-kernel.git] / fs / proc / proc_sysctl.c

/*
 * /proc/sys support
 */

#include <linux/sysctl.h>
#include <linux/proc_fs.h>
#include <linux/security.h>
#include "internal.h"

static struct dentry_operations proc_sys_dentry_operations;
static const struct file_operations proc_sys_file_operations;
static struct inode_operations proc_sys_inode_operations;

static void proc_sys_refresh_inode(struct inode *inode, struct ctl_table *table)
{
	/* Refresh the cached information bits in the inode */
	if (table) {
		inode->i_uid = 0;
		inode->i_gid = 0;
		inode->i_mode = table->mode;
		if (table->proc_handler) {
			inode->i_mode |= S_IFREG;
			inode->i_nlink = 1;
		} else {
			inode->i_mode |= S_IFDIR;
			inode->i_nlink = 0;	/* It is too hard to figure out */
		}
	}
}

static struct inode *proc_sys_make_inode(struct inode *dir, struct ctl_table *table)
{
	struct inode *inode;
	struct proc_inode *dir_ei, *ei;
	int depth;

	inode = new_inode(dir->i_sb);
	if (!inode)
		goto out;

	/* A directory is always one deeper than it's parent */
	dir_ei = PROC_I(dir);
	depth = dir_ei->fd + 1;

	ei = PROC_I(inode);
	ei->fd = depth;
	inode->i_mtime = inode->i_atime = inode->i_ctime = CURRENT_TIME;
	inode->i_op = &proc_sys_inode_operations;
	inode->i_fop = &proc_sys_file_operations;
	proc_sys_refresh_inode(inode, table);
out:
	return inode;
}

static struct dentry *proc_sys_ancestor(struct dentry *dentry, int depth)
{
	for (;;) {
		struct proc_inode *ei;

		ei = PROC_I(dentry->d_inode);
		if (ei->fd == depth)
			break; /* found */

		dentry = dentry->d_parent;
	}
	return dentry;
}

static struct ctl_table *proc_sys_lookup_table_one(struct ctl_table *table,
							struct qstr *name)
{
	int len;
	for ( ; table->ctl_name || table->procname; table++) {

		if (!table->procname)
			continue;

		len = strlen(table->procname);
		if (len != name->len)
			continue;

		if (memcmp(table->procname, name->name, len) != 0)
			continue;

		/* I have a match */
		return table;
	}
	return NULL;
}

static struct ctl_table *proc_sys_lookup_table(struct dentry *dentry,
						struct ctl_table *table)
{
	struct dentry *ancestor;
	struct proc_inode *ei;
	int depth, i;

	ei = PROC_I(dentry->d_inode);
	depth = ei->fd;

	if (depth == 0)
		return table;

	for (i = 1; table && (i <= depth); i++) {
		ancestor = proc_sys_ancestor(dentry, i);
		table = proc_sys_lookup_table_one(table, &ancestor->d_name);
		if (table)
			table = table->child;
	}
	return table;

}
static struct ctl_table *proc_sys_lookup_entry(struct dentry *dparent,
						struct qstr *name,
						struct ctl_table *table)
{
	table = proc_sys_lookup_table(dparent, table);
	if (table)
		table = proc_sys_lookup_table_one(table, name);
	return table;
}

static struct ctl_table *do_proc_sys_lookup(struct dentry *parent,
						struct qstr *name,
						struct ctl_table_header **ptr)
{
	struct ctl_table_header *head;
	struct ctl_table *table = NULL;

	for (head = sysctl_head_next(NULL); head;
			head = sysctl_head_next(head)) {
		table = proc_sys_lookup_entry(parent, name, head->ctl_table);
		if (table)
			break;
	}
	*ptr = head;
	return table;
}

static struct dentry *proc_sys_lookup(struct inode *dir, struct dentry *dentry,
					struct nameidata *nd)
{
	struct ctl_table_header *head;
	struct inode *inode;
	struct dentry *err;
	struct ctl_table *table;

	err = ERR_PTR(-ENOENT);
	table = do_proc_sys_lookup(dentry->d_parent, &dentry->d_name, &head);
	if (!table)
		goto out;

	err = ERR_PTR(-ENOMEM);
	inode = proc_sys_make_inode(dir, table);
	if (!inode)
		goto out;

	err = NULL;
	dentry->d_op = &proc_sys_dentry_operations;
	d_add(dentry, inode);

out:
	sysctl_head_finish(head);
	return err;
}

static ssize_t proc_sys_read(struct file *filp, char __user *buf,
				size_t count, loff_t *ppos)
{
	struct dentry *dentry = filp->f_dentry;
	struct ctl_table_header *head;
	struct ctl_table *table;
	ssize_t error, res;

	table = do_proc_sys_lookup(dentry->d_parent, &dentry->d_name, &head);
	/* Has the sysctl entry disappeared on us? */
	error = -ENOENT;
	if (!table)
		goto out;

	/* Has the sysctl entry been replaced by a directory? */
	error = -EISDIR;
	if (!table->proc_handler)
		goto out;

	/*
	 * At this point we know that the sysctl was not unregistered
	 * and won't be until we finish.
	 */
	error = -EPERM;
	if (sysctl_perm(table, MAY_READ))
		goto out;

	/* careful: calling conventions are nasty here */
	res = count;
	error = table->proc_handler(table, 0, filp, buf, &res, ppos);
	if (!error)
		error = res;
out:
	sysctl_head_finish(head);

	return error;
}

static ssize_t proc_sys_write(struct file *filp, const char __user *buf,
				size_t count, loff_t *ppos)
{
	struct dentry *dentry = filp->f_dentry;
	struct ctl_table_header *head;
	struct ctl_table *table;
	ssize_t error, res;

	table = do_proc_sys_lookup(dentry->d_parent, &dentry->d_name, &head);
	/* Has the sysctl entry disappeared on us? */
	error = -ENOENT;
	if (!table)
		goto out;

	/* Has the sysctl entry been replaced by a directory? */
	error = -EISDIR;
	if (!table->proc_handler)
		goto out;

	/*
	 * At this point we know that the sysctl was not unregistered
	 * and won't be until we finish.
	 */
	error = -EPERM;
	if (sysctl_perm(table, MAY_WRITE))
		goto out;

	/* careful: calling conventions are nasty here */
	res = count;
	error = table->proc_handler(table, 1, filp, (char __user *)buf,
				    &res, ppos);
	if (!error)
		error = res;
out:
	sysctl_head_finish(head);

	return error;
}


static int proc_sys_fill_cache(struct file *filp, void *dirent,
				filldir_t filldir, struct ctl_table *table)
{
	struct ctl_table_header *head;
	struct ctl_table *child_table = NULL;
	struct dentry *child, *dir = filp->f_path.dentry;
	struct inode *inode;
	struct qstr qname;
	ino_t ino = 0;
	unsigned type = DT_UNKNOWN;
	int ret;

	qname.name = table->procname;
	qname.len  = strlen(table->procname);
	qname.hash = full_name_hash(qname.name, qname.len);

	/* Suppress duplicates.
	 * Only fill a directory entry if it is the value that
	 * an ordinary lookup of that name returns.  Hide all
	 * others.
	 *
	 * If we ever cache this translation in the dcache
	 * I should do a dcache lookup first.  But for now
	 * it is just simpler not to.
	 */
	ret = 0;
	child_table = do_proc_sys_lookup(dir, &qname, &head);
	sysctl_head_finish(head);
	if (child_table != table)
		return 0;

	child = d_lookup(dir, &qname);
	if (!child) {
		struct dentry *new;
		new = d_alloc(dir, &qname);
		if (new) {
			inode = proc_sys_make_inode(dir->d_inode, table);
			if (!inode)
				child = ERR_PTR(-ENOMEM);
			else {
				new->d_op = &proc_sys_dentry_operations;
				d_add(new, inode);
			}
			if (child)
				dput(new);
			else
				child = new;
		}
	}
	if (!child || IS_ERR(child) || !child->d_inode)
		goto end_instantiate;
	inode = child->d_inode;
	if (inode) {
		ino  = inode->i_ino;
		type = inode->i_mode >> 12;
	}
	dput(child);
end_instantiate:
	if (!ino)
		ino= find_inode_number(dir, &qname);
	if (!ino)
		ino = 1;
	return filldir(dirent, qname.name, qname.len, filp->f_pos, ino, type);
}

static int proc_sys_readdir(struct file *filp, void *dirent, filldir_t filldir)
{
	struct dentry *dentry = filp->f_dentry;
	struct inode *inode = dentry->d_inode;
	struct ctl_table_header *head = NULL;
	struct ctl_table *table;
	unsigned long pos;
	int ret;

	ret = -ENOTDIR;
	if (!S_ISDIR(inode->i_mode))
		goto out;

	ret = 0;
	/* Avoid a switch here: arm builds fail with missing __cmpdi2 */
	if (filp->f_pos == 0) {
		if (filldir(dirent, ".", 1, filp->f_pos,
				inode->i_ino, DT_DIR) < 0)
			goto out;
		filp->f_pos++;
	}
	if (filp->f_pos == 1) {
		if (filldir(dirent, "..", 2, filp->f_pos,
				parent_ino(dentry), DT_DIR) < 0)
			goto out;
		filp->f_pos++;
	}
	pos = 2;

	/* - Find each instance of the directory
	 * - Read all entries in each instance
	 * - Before returning an entry to user space lookup the entry
	 *   by name and if I find a different entry don't return
	 *   this one because it means it is a buried dup.
	 * For sysctl this should only happen for directory entries.
	 */
	for (head = sysctl_head_next(NULL); head; head = sysctl_head_next(head)) {
		table = proc_sys_lookup_table(dentry, head->ctl_table);

		if (!table)
			continue;

		for (; table->ctl_name || table->procname; table++, pos++) {
			/* Can't do anything without a proc name */
			if (!table->procname)
				continue;

			if (pos < filp->f_pos)
				continue;

			if (proc_sys_fill_cache(filp, dirent, filldir, table) < 0)
				goto out;
			filp->f_pos = pos + 1;
		}
	}
	ret = 1;
out:
	sysctl_head_finish(head);
	return ret;
}

static int proc_sys_permission(struct inode *inode, int mask, struct nameidata *nd)
{
	/*
	 * sysctl entries that are not writeable,
	 * are _NOT_ writeable, capabilities or not.
	 */
	struct ctl_table_header *head;
	struct ctl_table *table;
	struct dentry *dentry;
	int mode;
	int depth;
	int error;

	head = NULL;
	depth = PROC_I(inode)->fd;

	/* First check the cached permissions, in case we don't have
	 * enough information to lookup the sysctl table entry.
	 */
	error = -EACCES;
	mode = inode->i_mode;

	if (current->euid == 0)
		mode >>= 6;
	else if (in_group_p(0))
		mode >>= 3;

	if ((mode & mask & (MAY_READ|MAY_WRITE|MAY_EXEC)) == mask)
		error = 0;

	/* If we can't get a sysctl table entry the permission
	 * checks on the cached mode will have to be enough.
	 */
	if (!nd || !depth)
		goto out;

	dentry = nd->dentry;
	table = do_proc_sys_lookup(dentry->d_parent, &dentry->d_name, &head);

	/* If the entry does not exist deny permission */
	error = -EACCES;
	if (!table)
		goto out;

	/* Use the permissions on the sysctl table entry */
	error = sysctl_perm(table, mask);
out:
	sysctl_head_finish(head);
	return error;
}

static int proc_sys_setattr(struct dentry *dentry, struct iattr *attr)
{
	struct inode *inode = dentry->d_inode;
	int error;

	if (attr->ia_valid & (ATTR_MODE | ATTR_UID | ATTR_GID))
		return -EPERM;

	error = inode_change_ok(inode, attr);
	if (!error) {
		error = security_inode_setattr(dentry, attr);
		if (!error)
			error = inode_setattr(inode, attr);
	}

	return error;
}

/* I'm lazy and don't distinguish between files and directories,
 * until access time.
 */
static const struct file_operations proc_sys_file_operations = {
	.read		= proc_sys_read,
	.write		= proc_sys_write,
	.readdir	= proc_sys_readdir,
};

static struct inode_operations proc_sys_inode_operations = {
	.lookup		= proc_sys_lookup,
	.permission	= proc_sys_permission,
	.setattr	= proc_sys_setattr,
};

static int proc_sys_revalidate(struct dentry *dentry, struct nameidata *nd)
{
	struct ctl_table_header *head;
	struct ctl_table *table;
	table = do_proc_sys_lookup(dentry->d_parent, &dentry->d_name, &head);
	proc_sys_refresh_inode(dentry->d_inode, table);
	sysctl_head_finish(head);
	return !!table;
}

static struct dentry_operations proc_sys_dentry_operations = {
	.d_revalidate	= proc_sys_revalidate,
};

static struct proc_dir_entry *proc_sys_root;

int proc_sys_init(void)
{
	proc_sys_root = proc_mkdir("sys", NULL);
	proc_sys_root->proc_iops = &proc_sys_inode_operations;
	proc_sys_root->proc_fops = &proc_sys_file_operations;
	proc_sys_root->nlink = 0;
	return 0;
}
Commit	Line	Data
77b14db5 EB	1	/*
	2	* /proc/sys support
	3	*/
	4
	5	#include <linux/sysctl.h>
	6	#include <linux/proc_fs.h>
	7	#include <linux/security.h>
	8	#include "internal.h"
	9
	10	static struct dentry_operations proc_sys_dentry_operations;
	11	static const struct file_operations proc_sys_file_operations;
	12	static struct inode_operations proc_sys_inode_operations;
	13
	14	static void proc_sys_refresh_inode(struct inode inode, struct ctl_table table)
	15	{
	16	/* Refresh the cached information bits in the inode */
	17	if (table) {
	18	inode->i_uid = 0;
	19	inode->i_gid = 0;
	20	inode->i_mode = table->mode;
	21	if (table->proc_handler) {
	22	inode->i_mode \|= S_IFREG;
	23	inode->i_nlink = 1;
	24	} else {
	25	inode->i_mode \|= S_IFDIR;
	26	inode->i_nlink = 0; /* It is too hard to figure out */
	27	}
	28	}
	29	}
	30
	31	static struct inode proc_sys_make_inode(struct inode dir, struct ctl_table *table)
	32	{
	33	struct inode *inode;
	34	struct proc_inode dir_ei, ei;
	35	int depth;
	36
	37	inode = new_inode(dir->i_sb);
	38	if (!inode)
	39	goto out;
	40
	41	/* A directory is always one deeper than it's parent */
	42	dir_ei = PROC_I(dir);
	43	depth = dir_ei->fd + 1;
	44
	45	ei = PROC_I(inode);
	46	ei->fd = depth;
	47	inode->i_mtime = inode->i_atime = inode->i_ctime = CURRENT_TIME;
	48	inode->i_op = &proc_sys_inode_operations;
	49	inode->i_fop = &proc_sys_file_operations;
	50	proc_sys_refresh_inode(inode, table);
	51	out:
	52	return inode;
	53	}
	54
	55	static struct dentry proc_sys_ancestor(struct dentry dentry, int depth)
	56	{
	57	for (;;) {
	58	struct proc_inode *ei;
	59
	60	ei = PROC_I(dentry->d_inode);
	61	if (ei->fd == depth)
	62	break; /* found */
	63
	64	dentry = dentry->d_parent;
65	}
66	return dentry;
67	}
68
69	static struct ctl_table proc_sys_lookup_table_one(struct ctl_table table,
70	struct qstr *name)
71	{
72	int len;
73	for ( ; table->ctl_name \|\| table->procname; table++) {
74
75	if (!table->procname)
76	continue;
77
78	len = strlen(table->procname);
79	if (len != name->len)
80	continue;
81
82	if (memcmp(table->procname, name->name, len) != 0)
83	continue;
84
85	/* I have a match */
86	return table;
87	}
88	return NULL;
89	}
90
91	static struct ctl_table proc_sys_lookup_table(struct dentry dentry,
92	struct ctl_table *table)
93	{
94	struct dentry *ancestor;
95	struct proc_inode *ei;
96	int depth, i;
97
98	ei = PROC_I(dentry->d_inode);
99	depth = ei->fd;
100
101	if (depth == 0)
102	return table;
103
104	for (i = 1; table && (i <= depth); i++) {
105	ancestor = proc_sys_ancestor(dentry, i);
106	table = proc_sys_lookup_table_one(table, &ancestor->d_name);
107	if (table)
108	table = table->child;
109	}
110	return table;
111
112	}
113	static struct ctl_table proc_sys_lookup_entry(struct dentry dparent,
114	struct qstr *name,
115	struct ctl_table *table)
116	{
117	table = proc_sys_lookup_table(dparent, table);
118	if (table)
119	table = proc_sys_lookup_table_one(table, name);
120	return table;
121	}
122
123	static struct ctl_table do_proc_sys_lookup(struct dentry parent,
124	struct qstr *name,
125	struct ctl_table_header **ptr)
126	{
127	struct ctl_table_header *head;
128	struct ctl_table *table = NULL;
129
130	for (head = sysctl_head_next(NULL); head;
131	head = sysctl_head_next(head)) {
132	table = proc_sys_lookup_entry(parent, name, head->ctl_table);
133	if (table)
134	break;
135	}
136	*ptr = head;
137	return table;
138	}
139
140	static struct dentry proc_sys_lookup(struct inode dir, struct dentry *dentry,
141	struct nameidata *nd)
142	{
143	struct ctl_table_header *head;
144	struct inode *inode;
145	struct dentry *err;
146	struct ctl_table *table;
147
148	err = ERR_PTR(-ENOENT);
149	table = do_proc_sys_lookup(dentry->d_parent, &dentry->d_name, &head);
150	if (!table)
151	goto out;
152
153	err = ERR_PTR(-ENOMEM);
154	inode = proc_sys_make_inode(dir, table);
155	if (!inode)
156	goto out;
157
158	err = NULL;
159	dentry->d_op = &proc_sys_dentry_operations;
160	d_add(dentry, inode);
161
162	out:
163	sysctl_head_finish(head);
164	return err;
165	}
166
167	static ssize_t proc_sys_read(struct file filp, char __user buf,
168	size_t count, loff_t *ppos)
169	{
170	struct dentry *dentry = filp->f_dentry;
171	struct ctl_table_header *head;
172	struct ctl_table *table;
173	ssize_t error, res;
174
175	table = do_proc_sys_lookup(dentry->d_parent, &dentry->d_name, &head);
176	/* Has the sysctl entry disappeared on us? */
177	error = -ENOENT;
178	if (!table)
179	goto out;
180
181	/* Has the sysctl entry been replaced by a directory? */
182	error = -EISDIR;
183	if (!table->proc_handler)
184	goto out;
185
186	/*
187	* At this point we know that the sysctl was not unregistered
188	* and won't be until we finish.
189	*/
190	error = -EPERM;
191	if (sysctl_perm(table, MAY_READ))
192	goto out;
193
194	/* careful: calling conventions are nasty here */
195	res = count;
196	error = table->proc_handler(table, 0, filp, buf, &res, ppos);
197	if (!error)
198	error = res;
199	out:
200	sysctl_head_finish(head);
201
202	return error;
203	}
204
205	static ssize_t proc_sys_write(struct file filp, const char __user buf,
206	size_t count, loff_t *ppos)
207	{
208	struct dentry *dentry = filp->f_dentry;
209	struct ctl_table_header *head;
210	struct ctl_table *table;
211	ssize_t error, res;
212
213	table = do_proc_sys_lookup(dentry->d_parent, &dentry->d_name, &head);
214	/* Has the sysctl entry disappeared on us? */
215	error = -ENOENT;
216	if (!table)
217	goto out;
218
219	/* Has the sysctl entry been replaced by a directory? */
220	error = -EISDIR;
221	if (!table->proc_handler)
222	goto out;
223
224	/*
225	* At this point we know that the sysctl was not unregistered
226	* and won't be until we finish.
227	*/
228	error = -EPERM;
229	if (sysctl_perm(table, MAY_WRITE))
230	goto out;
231
232	/* careful: calling conventions are nasty here */
233	res = count;
234	error = table->proc_handler(table, 1, filp, (char __user *)buf,
235	&res, ppos);
236	if (!error)
237	error = res;
238	out:
239	sysctl_head_finish(head);
240
241	return error;
242	}
243
244
245	static int proc_sys_fill_cache(struct file filp, void dirent,
246	filldir_t filldir, struct ctl_table *table)
247	{
248	struct ctl_table_header *head;
249	struct ctl_table *child_table = NULL;
250	struct dentry child, dir = filp->f_path.dentry;
251	struct inode *inode;
252	struct qstr qname;
253	ino_t ino = 0;
254	unsigned type = DT_UNKNOWN;
255	int ret;
256
257	qname.name = table->procname;
258	qname.len = strlen(table->procname);
259	qname.hash = full_name_hash(qname.name, qname.len);
260
261	/* Suppress duplicates.
262	* Only fill a directory entry if it is the value that
263	* an ordinary lookup of that name returns. Hide all
264	* others.
265	*
266	* If we ever cache this translation in the dcache
267	* I should do a dcache lookup first. But for now
268	* it is just simpler not to.
269	*/
270	ret = 0;
271	child_table = do_proc_sys_lookup(dir, &qname, &head);
272	sysctl_head_finish(head);
273	if (child_table != table)
274	return 0;
275
276	child = d_lookup(dir, &qname);
277	if (!child) {
278	struct dentry *new;
279	new = d_alloc(dir, &qname);
280	if (new) {
281	inode = proc_sys_make_inode(dir->d_inode, table);
282	if (!inode)
283	child = ERR_PTR(-ENOMEM);
284	else {
285	new->d_op = &proc_sys_dentry_operations;
286	d_add(new, inode);
287	}
288	if (child)
289	dput(new);
290	else
291	child = new;
292	}
293	}
294	if (!child \|\| IS_ERR(child) \|\| !child->d_inode)
295	goto end_instantiate;
296	inode = child->d_inode;
297	if (inode) {
298	ino = inode->i_ino;
299	type = inode->i_mode >> 12;
300	}
301	dput(child);
302	end_instantiate:
303	if (!ino)
304	ino= find_inode_number(dir, &qname);
305	if (!ino)
306	ino = 1;
307	return filldir(dirent, qname.name, qname.len, filp->f_pos, ino, type);
308	}
309
310	static int proc_sys_readdir(struct file filp, void dirent, filldir_t filldir)
311	{
312	struct dentry *dentry = filp->f_dentry;
313	struct inode *inode = dentry->d_inode;
314	struct ctl_table_header *head = NULL;
315	struct ctl_table *table;
316	unsigned long pos;
317	int ret;
318
319	ret = -ENOTDIR;
320	if (!S_ISDIR(inode->i_mode))
321	goto out;
322
323	ret = 0;
324	/* Avoid a switch here: arm builds fail with missing __cmpdi2 */
325	if (filp->f_pos == 0) {
326	if (filldir(dirent, ".", 1, filp->f_pos,
327	inode->i_ino, DT_DIR) < 0)
328	goto out;
329	filp->f_pos++;
330	}
331	if (filp->f_pos == 1) {
332	if (filldir(dirent, "..", 2, filp->f_pos,
333	parent_ino(dentry), DT_DIR) < 0)
334	goto out;
335	filp->f_pos++;
336	}
337	pos = 2;
338
339	/* - Find each instance of the directory
340	* - Read all entries in each instance
341	* - Before returning an entry to user space lookup the entry
342	* by name and if I find a different entry don't return
343	* this one because it means it is a buried dup.
344	* For sysctl this should only happen for directory entries.
345	*/
346	for (head = sysctl_head_next(NULL); head; head = sysctl_head_next(head)) {
347	table = proc_sys_lookup_table(dentry, head->ctl_table);
348
349	if (!table)
350	continue;
351
352	for (; table->ctl_name \|\| table->procname; table++, pos++) {
353	/* Can't do anything without a proc name */
354	if (!table->procname)
355	continue;
356
357	if (pos < filp->f_pos)
358	continue;
359
360	if (proc_sys_fill_cache(filp, dirent, filldir, table) < 0)
361	goto out;
362	filp->f_pos = pos + 1;
363	}
364	}
365	ret = 1;
366	out:
367	sysctl_head_finish(head);
368	return ret;
369	}
370
371	static int proc_sys_permission(struct inode inode, int mask, struct nameidata nd)
372	{
373	/*
374	* sysctl entries that are not writeable,
375	* are _NOT_ writeable, capabilities or not.
376	*/
377	struct ctl_table_header *head;
378	struct ctl_table *table;
379	struct dentry *dentry;
380	int mode;
381	int depth;
382	int error;
383
384	head = NULL;
385	depth = PROC_I(inode)->fd;
386
387	/* First check the cached permissions, in case we don't have
388	* enough information to lookup the sysctl table entry.
389	*/
390	error = -EACCES;
391	mode = inode->i_mode;
392
393	if (current->euid == 0)
394	mode >>= 6;
395	else if (in_group_p(0))
396	mode >>= 3;
397
398	if ((mode & mask & (MAY_READ\|MAY_WRITE\|MAY_EXEC)) == mask)
399	error = 0;
400
401	/* If we can't get a sysctl table entry the permission
402	* checks on the cached mode will have to be enough.
403	*/
404	if (!nd \|\| !depth)
405	goto out;
406
407	dentry = nd->dentry;
408	table = do_proc_sys_lookup(dentry->d_parent, &dentry->d_name, &head);
409
410	/* If the entry does not exist deny permission */
411	error = -EACCES;
412	if (!table)
413	goto out;
414
415	/* Use the permissions on the sysctl table entry */
416	error = sysctl_perm(table, mask);
417	out:
418	sysctl_head_finish(head);
419	return error;
420	}
421
422	static int proc_sys_setattr(struct dentry dentry, struct iattr attr)
423	{
424	struct inode *inode = dentry->d_inode;
425	int error;
426
427	if (attr->ia_valid & (ATTR_MODE \| ATTR_UID \| ATTR_GID))
428	return -EPERM;
429
430	error = inode_change_ok(inode, attr);
431	if (!error) {
432	error = security_inode_setattr(dentry, attr);
433	if (!error)
434	error = inode_setattr(inode, attr);
435	}
436
437	return error;
438	}
439
440	/* I'm lazy and don't distinguish between files and directories,
441	* until access time.
442	*/
443	static const struct file_operations proc_sys_file_operations = {
444	.read = proc_sys_read,
445	.write = proc_sys_write,
446	.readdir = proc_sys_readdir,
447	};
448
449	static struct inode_operations proc_sys_inode_operations = {
450	.lookup = proc_sys_lookup,
451	.permission = proc_sys_permission,
452	.setattr = proc_sys_setattr,
453	};
454
455	static int proc_sys_revalidate(struct dentry dentry, struct nameidata nd)
456	{
457	struct ctl_table_header *head;
458	struct ctl_table *table;
459	table = do_proc_sys_lookup(dentry->d_parent, &dentry->d_name, &head);
460	proc_sys_refresh_inode(dentry->d_inode, table);
461	sysctl_head_finish(head);
462	return !!table;
463	}
464
465	static struct dentry_operations proc_sys_dentry_operations = {
466	.d_revalidate = proc_sys_revalidate,
467	};
468
469	static struct proc_dir_entry *proc_sys_root;
470
471	int proc_sys_init(void)
472	{
473	proc_sys_root = proc_mkdir("sys", NULL);
474	proc_sys_root->proc_iops = &proc_sys_inode_operations;
475	proc_sys_root->proc_fops = &proc_sys_file_operations;
476	proc_sys_root->nlink = 0;
477	return 0;
478	}